The main motivation of this patch was to merge the three initialization
functions in one and clean it up. However as some changes in other code
areas where needed to do it right some small other changes were made.
Changes to viafb_par:
io_virt renamed as engine_mmio and moved to shared
VQ_start renamed as vq_vram_addr and moved to shared
VQ_end removed as it is easily recalculatable
vq_vram_addr is not strictly needed but keep it to track where we
allocated video memory. The memory allocated for the virtual queue was
shrunk to VQ_SIZE as VQ_SIZE+CURSOR_SIZE looked like a bug to me. But to
be honest I don't have the faintest idea what virtual queues are for in
the graphic hardware and whether the driver needs them in any way. I only
know that they aren't directly accessed by the driver and so the only
potential current use would be as hardware internal buffers. For now keep
them to avoid regressions and only remove the double cursor allocation.
The most changes were caused by renames and the mentioned structure
changes so the chance of regressions is pretty low. The meaning of
viafb_accel changed slightly as previously it was changed back and forth
in the code and allowed to enable the hardware acceleration by software if
previously disabled. The new behaviour is that viafb_accel=0 always
prevents hardware acceleration. With viafb_accel!=0 the acceleration can
be freely choosen by set_var. This means viafb_accel is a diagnostic tool
and if someone has to use viafb_accel=0 the driver needs to be fixed.
As this is mostly a code cleanup no regressions beside the slightly change
of viafb_accel is expected.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Split the pitch handling up and replaces the calculation from virtual xres
and bpp with fix.line_length which already contains the pitch and does not
add any constrains for the virtual resolution.
Also add a bit to the second pitch which the documentation mentions but
which was ignored by the driver.
Although it is a bit unclear what the right pitch for some LCD modes is
this patch should have no negative runtime impact.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Clean the hardware cursor handling up.
The most notable change is that it no longer buffers the values in
viacursor but uses the ones in cursor instead as they are guaranteed to be
always valid.
Furthermore it uses local instead global variables where possible, moves
the cursor variable in shared as only one hardware cursor is supported and
returns an error if memory allocation fails. Last but not least it fixes
a too small buffer (as u32 has only 4 and not 32 bytes) but this did not
produce any known problems.
This is mostly a code cleanup, no negative runtime changes are expected.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch is a completly rewritten 2D engine. The engine is no longer in
a default state but reinitialized every time to allow usage for both
framebuffers regardless of their settings.
The whole engine handling is concentrated in a big function which takes 16
parameters. Although the number of parameters is worryingly it is good to
have a single funtion to deal with this stuff as it allows to easily
support different engines and avoids some code duplication.
On the way support for the new 2D engine in VX800 was added. As the with
less code duplication but it is probably better to duplicate the code as
this way is easier to walk if VIA ever decides to release a new engine
which changes anything the driver touches.
The engine support for VX800 gives a notable boost in speed. There are no
known regressions but as this patch changes paths I do neither have the
hardware nor documentation to check and has the possibility to put the
system in a critical state heavy testing is appreciated.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch introduces viafb_shared and is the beginning of a smooth
transition to use it.
viafb_shared should contain all general, non-surface specific data that
should be shared along all viafb framebuffers while viafb_par should only
contain things that are specific to each surface or in other words extend
fb_info. This change is intended to clean the dual/multi framebuffer
handling up.
This removes the annoyance that viafbinfo1->par points to a different
structure than viaparinfo1.
As the last change is fundamental it is difficult to ensure that all parts
of the driver do not depend on the previous brokenness but the chance of
regressions is very low.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This removes the completly useless io variable as well as the temporary
used variables mmio_base and mmio_len in favor to use directly the fb_info
variables.
This is a code cleanup only, no runtime change expected.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Remove everything related to video devices from the driver as it did not
influence the driver operation. This patch does change the userspace
behaviour as it removes two IOCTLs and one module parameter. But this is
good as it removes useless stuff and helps the user to figure out the
options that do affect the driver behaviour (which are still too many).
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Clean the handling of ioremapped video memory up. The following changes
were made:
info->screen_base - viafb_FB_MM
(VRAM offset calculation) was replaced by
info->fix.smem_start - viafbinfo->fix.smem_start
which is essentially the same calculation but done with physical instead
virtual addresses.
*->fbmem_virt
was replaced by
viafbinfo->screen_base
This is true for viafbinfo and viafbinfo1 as the par pointers are equal.
An early initialization of viafbinfo1->fix.smem* was removed as done later
in viafb_setup_fixinfo.
This patch highlights that the only usage of the ioremapped video memory
in the driver is for hardware cursor handling. Even if it has to hold the
used virtual screen mapped for old-fashioned read/write calls (vs.
mmap'ed) a lot virtual memory could be saved by only ioremapping on
demand.
Code cleanup, no runtime changes expected.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Clean the duoview handling up by replacing the varible with the funtion in
the only place where it is used. This is a code cleanup only, no runtime
change expected.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Remove the mode information from viafbdev.c and uses the one of viamode.c
instead. This is possible because horizontal and vertical address are the
same as horizontal and vertical resolution. The reduced blanking modes in
the table are no problem because they have a higher index than the normal
modes and therefore always the normal modes are selected just as the old
behaviour.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Shrink and merge viafb_update_viafb_par. This removes a lot of duplicated
data in viafb_par. Use the relevant data of fb_info instead. On the way
it removes an inconsistency in handling a second framebuffer which only
worked because viafbinfo1->par is modified to point to the same viafb_par
as viafbinfo->par.
Code cleanup only, no runtime change expected.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Cc: Harald Welte <laforge@gnumonks.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This fixes a bug caused by changing pointers (viafb_mode, viafb_mode1)
assigned by module_param. It reduces driver complexity by not needlessly
changing these vars as they are only read once and removing now
superfluous code.
On unpatched kernels loading viafb with viafb_mode or viafb_mode1 option
used and afterwards unloading it results in:
kernel BUG at mm/slub.c:2926!
invalid opcode: 0000 [#1] PREEMPT
last sysfs file: /sys/devices/virtual/block/loop0/removable
Modules linked in: snd_hda_codec_realtek snd_hda_intel snd_hda_codec
snd_hwdep snd_pcm rtl8187 snd_timer eeprom_93cx6 mmc_block snd soundcore
via_sdmmc fb snd_page_alloc i2c_algo_bit i2c_viapro ehci_hcd uhci_hcd
cfbcopyarea mmc_core cfbimgblt cfbfillrect video output [last unloaded:
viafb]
Pid: 3355, comm: rmmod Not tainted (2.6.31-rc1 #0)
EIP: 0060:[<c106a759>] EFLAGS: 00010246 CPU: 0
EIP is at kfree+0x80/0xda
EAX: c17c2da0 EBX: dc7edbdc ECX: 0000010f EDX: 00000000
ESI: c102c700 EDI: dc7ed8fa EBP: d703ff2c ESP: d703ff20
DS: 007b ES: 007b FS: 0000 GS: 0033 SS: 0068
Process rmmod (pid: 3355, ti=d703e000 task=db1412c0 task.ti=d703e000)
Stack:
dc7edbdc 00000014 00000016 d703ff40 c102c700 dc7f45d4 dc7f45d4 00000880
d703ff4c c103e571 00000000 d703ffac c103e751 66616976 da140062 db89ba80
00000328 d702edf8 db89ba80 d703ff9c c105d0f0 00000200 da14f898 00000014
Call Trace:
[<c102c700>] ? destroy_params+0x1e/0x2b
[<c103e571>] ? free_module+0xa2/0xd7
[<c103e751>] ? sys_delete_module+0x1ab/0x1da
[<c105d0f0>] ? do_munmap+0x20a/0x225
[<c10029b4>] ? sysenter_do_call+0x12/0x26
Code: 10 76 7a 8d 87 00 00 00 40 c1 e8 0c c1 e0 05 03 05 1c 87 41 c1 66 83 38 00 79 03 8b 40 0c 8b 10 84 d2 78 12 66 f7 c2 00 c0 75 04 <0f> 0b eb fe e8 6f 5a fe ff eb 47 8b 55 04 8b 58 0c 9c 5e fa 3b
EIP: [<c106a759>] kfree+0x80/0xda SS:ESP 0068:d703ff20
This is caused by the current code changing the pointers assigned by
module_param. During unload it tries to free the memory the pointers
point at which is now part of an internal structure.
The patch simply avoids changing the pointers. This is okay as they are
read only once during the initialization process.
Signed-off-by: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Scott Fang <ScottFang@viatech.com.cn>
Cc: Joseph Chan <JosephChan@via.com.tw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Correct by following Jiri Slaby's suggestions.
Initialization, remove and some other important functions of viafb.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Joseph Chan <josephchan@via.com.tw>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
